Being less restrictive on numpy and pandas dependencies

[tarun292]

🚀 The feature, motivation and pitch

Currently in our pyproject.toml we're hardcoding to a certain version of numpy and pandas. These libraries are very commonly used in other model repos and there's a high probability that this will already be installed on a user's machine when they try to install ExecuTorch.

Can we make our numpy dependency less restrictive and instead limit numpy compatibility to a range of versions e.g. numpy>=2.1, numpy<3? Same with the pandas dependency. This will ensure that we don't cause dependency conflicts when users install ExecuTorch.

Alternatives

No response

Additional context

No response

RFC (Optional)

No response